In this article we discuss the use of big corpuses or databases as a first step for qualitative analysis of linguistic data. We concentrate on ASIt, the Syntactic Atlas of Italy, and take into consideration the different types of dialectal data that can be collected from similar corpora and databases. We analyze all the methodological problems derived from the necessary compromise between the strict requirements imposed by a scientific inquiry and the management of big amounts of data. As a possible solution, we propose that the type of variation is per se a tool to derive meaningful generalizations. To implement this idea, we examine three different types of variation patterns that can be used in the study of morpho-syntax: the geographical distribution of properties (and their total or partial overlapping, or complementary distribution), the so-called leopard spots variation, and the lexical variation index, which can be used to determine the internal complexity of functional items.